Enter Note Done

Session


Keywords: Clinical testing; Diagnostics; Exome sequencing

Authors:
K. Dumas 1; D. Fisk 1; M. Grove 1; C. Reavey 1; S. White 1; J. Buchan 1,2; E. Spiteri 1,2

Affiliations:
1) Stanford Health Care, Stanford, California; 2) Stanford University, Stanford, California


Next generation sequencing technologies provide a means to detect vast quantities of rare variation in clinical subjects. With this opportunity comes the challenge of distinguishing reportable, clinically-relevant variation from thousands of non-reportable variants. Clinical laboratories have developed a variety of methods for prioritizing variants for manual review, including the use of patient phenotype information to create targeted gene lists that prioritize variants within genes potentially related to patient disease phenotypes. Using patient medical records and HPO terms listed by ordering providers, Stanford Medicine Clinical Genomics Program (CGP) generates a unique, phenotype-driven, target gene list (TGL) for patients referred to CGP for clinical exome analysis. This phenotype-driven prioritization approach is run upstream of broader variant prioritization strategies to maximize the chances of identifying clinically-relevant variation. Here, we retrospectively investigate the value of a phenotype-driven variant prioritization strategy on a cohort of proband-only and trio samples referred to CGP for clinical exome sequencing and analysis.
Gene, variant and prioritization data were collected from each case, including: number of genes on each patient’s TGL, number of variants prioritized through the TGL, total number of prioritized variants, report result, and reported variant info. We investigated whether reported variants were identified via target gene strategies or broader strategies, whether variants identified by TGLs would have been identified by broader strategies had TGLs not been used, and which filters were most useful to prioritize variants when TGLs were not used.
Of the variants reported to clinicians, a majority were identified via the TGL, despite comprising a small quantity of the total prioritized variant list. For positive cases, about half of the reported variants were identified via the TGL. We observed that nearly all of our reported variants would have been identified by broader variant prioritization strategies had we not created a TGL. When TGLs were not used, inheritance and canonical loss of function filters were most useful in identifying reportable variants. Variants not detected in the absence of a TGL included rare, not previously reported, missense variation. These results demonstrate additional disease-gene association resources may help to improve phenotype-driven analysis.